Can Work-Family Policies reduce the poverty gap between one-parent and two-parent households?

Page created by Philip Wallace
 
CONTINUE READING
Can Work-Family Policies reduce the poverty gap between
one-parent and two-parent households?
- A Multi-Level Analysis of Child Poverty across 25 European Countries

Elisabeth Ugreninov, Bjørn Hvinden and Viggo Nordvik1

Introduction
Work-family policies are one of the most important welfare state innovations of the late
twentieth and early twenty-first centuries. The main objectives of these policies are to help
parents balancing competing demands of work and family and protect families against
poverty risk due to care demands. While the poverty risk varies cross-country and between
types of household, one-parent households have greater exposure of poverty than most other
types of households. Earlier research has attributed variations in the high poverty risk of one-
parent households, especially lone-mothers, to differences in employment rates and varieties
in the systems of social protection and work-family policies (e.g. Misra, Moller, and Budig
2007). In general, benefits are important to extend the total income for households with
children, and potentially pushing some vulnerable households above the poverty threshold,
thus also reducing their poverty risk. Nevertheless, benefits are seldom enough to keep
vulnerable households with children out of poverty. Similarly, employment is generally an
important factor of the reducing poverty risk. Although work-family policies aim to facilitate
the interplay between work and care obligations, some parents might have to reduce their
employment in periods to cope with their care burden, consequently being at greater poverty
risk. In this regard, accessible and affordable public or publically supported childcare is likely
to be an important factor. At the Barcelona Summit in 2002 as part of the European
Employment Strategy (European Council 2002), the Council agreed on an explicit target for
Member States to by 2010 provide childcare to at least 33% of children under 3 years old, and
to at least 90% of children between 3 and mandatory school age. Similarly, scholars have
defined investment in public childcare and commitment to equal distribution of public
childcare among all social groups, as an important steps on the way to reducing the child

1
 We are grateful for the many helpful comments colleagues in the COPE project have given to an earlier version
of this paper.

                                                                                                            1
poverty (e.g. Esping-Andersen et al. 2002). These scholars have pointed to at least two
reasons for expecting investment in public childcare to mitigate poverty. First, parents,
especially mothers, can step up their employment after birth, with the prospects of raising the
household income (the focus in this article). Second, such investment will enhance the human
capital of children by providing equal educational stimulation at an early stage, which in
particular would benefit children with disadvantaged background.

How the roles of parenting are understood, (e.g. whether mothers are regarded as the main
caregiver and fathers as the breadwinner, or not), are related to gendered views on the
appropriate interplay between families, state and the market (Daly and Lewis 2000;
O’Connor, Orloff and Shaver 1999). A basic dichotomy divides care responsibilities into
private and state. According to one view, care for children and family is a private and usually
women’s responsibility, meaning that the state has limited liability for the providing care
arrangements and cash transfers helping with the material aspects of upbringing children. By
a contrasting view, mothers have individual rights to uphold employment after birth, enabled
through the provision of paid parental leave and public childcare services reconciling the
work-family conflict. Nevertheless, this clear-cut dichotomy of views is weakened and
outdated. Nowadays, European welfare states aim to increase employment for both men and
women, and to facilitate dual earnership and contribute to gender equality in labour market
(Ghysels and Van Lancker 2011; Lewis et al. 2008; O’Connor 2005). Consequently, all
European welfare states utilize a mixture of universal and mean-tested cash benefits, tax
allowances and credits, and services, when supporting families with children. Nevertheless,
we still find substantial cross-country differences in the structure, generosity, and entitlement
firmly (Ferrarini 2006; Bradshaw 2006; Montanari 2000).

Given that the performance of work-family policies influences the risk of living in poverty,
two main questions arises: can work-family policies reduce the poverty gap between one-
parent and two-parent households? And, can work-family policies reduce the gender gap in
poverty among single-mothers and single-father households? The main aim of this article is to
investigate which configuration of work-family policies produces the best outcomes in terms
of households’ poverty risk and, in particular, if work-family policies can reduce the poverty
gap between one-parent and two-parent households, when taking into account the structures
of labour markets. Investigating this issue, we focus on key elements of work-family-related
policies, and assess their impact on poverty risks among different type of households with
children.

Our main attribution to the literature is to extend the understanding how the interplay between
work-family policies and one-parent households impacts poverty risk in a multilevel
perspective. While earlier research to a larger degree has solely focused on lone mothers, this

                                                                                               2
study compares one-parent with two-parent households on the one hand, and lone mothers
compared to lone fathers on the other hand. This extension is important to make progress in
our understanding of variations in poverty risks of households. In particular, such extension is
necessary to clarify whether differences in risks are driven by absence of the additional
income that a partner can bring (comparing one-parent with two-parent households), versus
by lack of resources and difficulties in the labour market (comparing single-mother with
single-father households). Hence this clarification includes a gender aspect as well.
Comparing one-parent and two-parent households will narrow the focus to households with
children.

Earlier research has mainly examined how different grouping of countries, conceptualised as
welfare regimes or care regimes, affects poverty, or has almost exclusively analysed poverty
risks at a macro or a micro-level approach. Typical macro-level studies assess the effects of
cross-national differences in social policy and cannot incorporate individual-level information
such as the characteristics of households with children. Vice versa, individual-level analyses
within countries using precise data at individual level, cannot grasp the impact of country
differences in policy or labour market situation. Recently, multilevel analyses have become
very popular. Multilevel analyses are essential when examining cross-national differences in
households’ poverty risk. This study joins this trend by examining the influence of household
and country characteristics on poverty risk using a multilevel analysis across 25 European
countries.

Earlier research and theoretical framework
When studying households’ poverty risk, research has to deal with two dimensions. First,
poverty may be a result of factors operating at the individual level, such as educational
attainment, employment, and household structure (the individual dimension). Second, a
number of societal factors may influence poverty risk, for instance a country’s work-family
policies, the level and structure of labour force participation, the availability of family and
other benefits and services, and overall affluence, for instance measured by the level or
change in GDP per capita. As we have mentioned earlier, earlier research largely tends to be
either macro- or micro-level studies. There is, however, a need to link individual and
household characteristics and societal factors in multilevel analyses, allowing us to examine
both micro- and macro-level explanations.

Starting with micro-level explanations, work and family are not separate spheres but influence
one another in a myriad of positive and negative ways (Campione 2008; Haddock et al. 2006;
Huang et al. 2004; Schieman and Young 2011; Tuttle and Garr 2009; Winslow 2005). First,
the relations between family structure and poverty risk are particularly prominent. Existing
scholarship suggests that the conflict and burden combining work and family are greater for

                                                                                              3
one-parent compared to two-parent households (Forma 2009; Winslow 2005). One-parent
households do face a greater burden in combining paid work and care responsibilities. Their
difficulties with combining work and family life affect the ability to obtain and maintain
employment (e.g. Ciabattari 2007; DeBord et al. 2000; Gemelli 2008; Howe and Pidwell
2004; Neblett 2007). Second, human capital differences between the heads of two-parent and
one-parent households, as well as between the heads of single-mother and single-father
households might affect their income earning capacity and thus their poverty risk. Often, one-
parent households, especially female headed, are more likely to have low income, and more
likely to report financial worries, suggesting that such families have less monetary resources
(Edin and Kefalas 2005; Malone et al. 2010; Son and Bauer 2010). Third, work place
organisations are in many societies more likely to fit workers who have a partner available to
attend non-work needs (Mannis 1999; Moen and Roehling 2005). For instance the employer
might think it is easier for an employee to work overtime if a partner can take care of the
children and domestic work. One-parent households’ greater needs for family-friendly work
place arrangements might cause a selection into apparently family-friendly but often low-
wage jobs.

Taking one-parents’ burden of combining work and family life, lack of human capital, and
poor family-friendly workplace organisation into account, we expect that one-parent
households have higher poverty risk than two-parent households regardless of other
households characteristics such as educational attainment, employment status, and age. We
suggest as well that single-mother households have a higher poverty risk than single-father
household.

The explanatory mechanisms we have mentioned affect households’ poverty risk differently
depending on cross-country differences of work-family policies and country structure. A
prominent theory used when comparing western welfare states is Esping-Andersen’s (1990)
threefold model of different welfare state regimes, each having its own features and
administration of risk developed. First, the liberal model (exemplified by the UK) is
characterised by the importance of the market and residual, mean-tested policies targeted at
specific groups. Second, the corporate-conservative model (exemplified by Germany, France,
Belgium, Netherlands, and Italy) is characterised by key role of the one-breadwinner family.
There is a strong connection between access to welfare services from the state and labour
force participation. Employment relationships assuming stable, non-interrupted working
careers serve to pool social risks. The state provides services only when the family, and not
the market, fails to produce the welfare needed. Third, the social-democratic model
(exemplified by the Nordic countries) is characterised by a state, securing access to income
maintenance and services which has a central and universal role in securing welfare. The
market and the family have marginal roles compared to the other welfare states.

                                                                                            4
A criticism, mainly among feminists, is that this welfare state typology is gender blind, and
that using different care regimes would strengthen the gender aspect (Lewis 1992; Orloff
1993). Drawing on Fraser’s (1994) conceptualization, four strategies for reconciling work and
family emerges. First, the care strategy, characterises women primarily as carers, and
combines a male breadwinner and a female carer strategy. The policy of carer countries (as
Netherlands, Germany and Luxemburg) explicitly rewards mothers for providing care,
reinforcing traditional gender divisions between care and employment. Part-time employment
is as an ideal strategy for mothers who wish to combine employment and care. Second, the
earner strategy, primarily treats women as workers and secondly as carers. Men and women
invest equally in labour force participation, but with low public support of care. The UK is a
case in point for this strategy. Third, the choice strategy, especially mothers can choose
whether they would like to be primarily earners or caregivers. Mothers are rewarded for
providing care but also stimulated to participate in the labour force. France and Belgium are
close to this mixed strategy. Policies provide substantial support both for mothers’ full-time
employment and for caregiving within the family. Fourth, the earner-carer strategy explicitly
has a gendered vision of society in which both mothers and fathers balance informal care and
employment. Mothers and fathers are stimulated to take parental leave, and after the leave,
have strong incentives and norms to re-enter employment. This is possible because available
public childcare after the leave period exist. Additional, employers meet the expectation to
accommodate the needs for parents to reduce the burden of combining work and informal
care. To link these care strategies to the poverty risk, earlier research indicates that the risk of
poverty is much lower in countries with the earner-care strategies, in general (Misra et al.
2007). In addition, poverty rates are significantly higher for one-parents, in particular, in
countries categorised as earner or care strategies (Misra et al. 2007).

Despite the heuristic value of such typologies of welfare or care regimes, for our purpose they
simplify matters too much to serve as a basis for formulating precise hypotheses for testing.
These and other existing typologies miss the complexity of factors affecting the poverty risks
of different combinations of household types and structural contexts. In other words, we argue
that work-family policies and structural contexts have different impact for different subgroups
of households. First, public childcare aims to reduce the burden of combining work and
family life and to increase especially mothers’ labour force participation (Pettit and Hook
2005), and through these mechanisms to decrease the poverty risk among household with
children. The effect of access to public childcare is, however, likely to be more complicated.
Such access will only reduce the poverty risk if marginal groups with high poverty risk use
public childcare to increase their employment. Nevertheless, existing research shows that
vulnerable groups tend not to use public childcare (Van Lanker 2012), thus undermining the
effect of public childcare on the poverty risk. Second, to have or get a job with a reasonable
wage may well be the most important way to avoid poverty. Vulnerable groups, in

                                                                                                  5
particularly one-parent households, tend, however, when employed to be or become low-
earners. Higher levels of employment for these groups are insufficient to lower their poverty
risk. Consequently, the availability and generosity of social transfers become crucial
(Christopher 2002a, 2002b, 2005; Christopher et al. 2002). Diverse forms of social transfers,
including family benefits, social assistance, minimum income and unemployment benefits, are
likely to decrease the poverty risk among households with children, independently of their
labour force participation.

Following Casper, McLanahan, and Garfinkel (1994) we ask whether low rates of
employment and one-parenthood are more significant than low human capital in explaining
poverty gaps, especially gender gaps in poverty. Country specific characteristics like overall
employment levels and generosity of social transfers are indeed important for the poverty risk
in general. However, do country specific characteristics give a better explanation than human
capital factors of the poverty gap between one-parent and two-parent households?
Additionally, how do we account for the gender gap in poverty between single-mother and
single-father households? To get a better understanding of the complexity of the interplay
between individual and country-specific characteristics, we rely on more fine-grained
measures at both micro- and macro-level. A better understanding of this interplay would
enable more reliable indicators of policy effects and more solid policy implications. Given
these considerations, there is a clear need for research that combines micro- and macro-levels
to examine the interplay between individual characteristics and country specific
characteristics on poverty risk among families with children.

The prevalence of poverty among the target groups at country level
To examine the statement that one-parents are vulnerable in European countries, we report
poverty figures for one-parents and two-parent households.

                                                                                            6
Figure 1. Poverty rates for lone parents and couples at active age (25-60) with dependent
children, European countries.

 70

 60

 50

 40
                                                                                    Two-parents

 30                                                                                 Lone-parents

 20

 10

  0
      RO ES PL LU LV LT BG IT PT MTHU UK EE SK BE FR DE AT SI FI SE CZ NO NL DK

Two observations emerge from this picture. First, in all countries, the results indicate that one-
parent households face a significantly higher poverty risk than two-parent households.
Second, cross-country differences are ranging from less than 10 per cent in Denmark to 59 per
cent in Malta. In-between cross-country variation (measured in percentage point) between
one-parent and two-parent households shows that country differences are greater in countries
where the overall poverty rate are high (Appendix 1). There is however, some exceptions,
notably Nordic countries as Sweden and Norway, reporting higher than expected poverty
figures among one-parent households. For instance, the gap between two-parent households’
poverty rates and one-parent households’ poverty rates are respectively 24 and 19 percentage
points. In contrast, the gap in Denmark is only 7 percentage points, and compared to other
countries with a high prevalence of poverty, for example Spain, the gap is 14 percentage
points. Earlier research found similar figures; see for example Van Lancker, Ghysels and
Cantillon (2012).

                                                                                                   7
The descriptive statistics indicate that one-parent households are associated with a higher
poverty risk vis-á-vis two-parent households. On the basis of previous research we have
highlighted several factors that may contribute to this result: age, educational attainment, and
labour force participation. Institutional configurations may mitigate or exacerbate the effect of
these determinants. To examine whether the association between high poverty risk and one-
parenthood at micro-level is spurious or not, we rely on multivariate analysis to be able to
control for competing explanations.

Empirical strategy
The dataset used is Survey on Living and Income Conditions (EU-SILC). EU-SILC provides
cross-nationally and representative micro-level dataset for all individuals within a household,
and provides comparable data on income, employment and living conditions of European
households. The dependent variable is a binary indicator of poverty, coded 1 for households
living in poverty and 0 for not having a poverty risk. We follow the vast majority of cross-
national poverty studies and use a relative measure of poverty. We consider a household as
poor if the equivalent household income is below 60% of the national median equivalent
household income in the country of residence. Household income is a function of income of
the members of the household. Multiple members are likely to pool their resources and
expenses, thus the unit of analysis are the household.

Sample restrictions
We conducted original analyses with a recent dataset for 25 European countries2. Our
analyses are confined to families with children, where a child is defined as a person below 18
years or below 24 years when economically inactive and living in the household. Further, we
restrict our sample to working-aged head adults, defined as those 25 through 60 years of age.
Partly, this is because concentrating on the working-aged, these age restrictions allow us to
hold constant some of the life cycle dynamics that increase the odds of being out of the labour
force. The analysis pools the data from these 25 countries into one merged files containing
67592 households.

Micro-level variables
At the micro-level, we consider several independent variables that are likely to affect poverty.
Four measures of head of household characteristics are included. Age and age square is a
scaled variable that ranges from 25 and onwards until 60 and refers to the age of the head of
the household. Age square is included to inspect whether the effect of age on the poverty risk
is increasing or decreasing after age. We categorize head educational attainment in three

2
  Austria, Belgium, Bulgaria, Czech republic, Deutschland, Denmark, Estonia, Spain, Finland, France, Hungary,
Italia, Lithuania, Luxemburg, Latvia, Malta, Netherland, Norway, Poland, Romania, Sweden, Slovenia, Slovak
republic, United Kingdom.

                                                                                                           8
different dummies based on the highest level of education of head using ISCED standard. If
the ISCED scale is used as a continuous variable, we thus predict that a change from level 1 to
2 has the same effect as a change from 4 to 5; binary variables are used to avoid such
problems. We include binary measures of Head low education (ISCED=0-2), Head medium
education (ISCED=3-4) and Head high education (ISCED=5).

Three measures of household characteristics are included. Type of household are organised as
a dummy, coded 0 if the household consists of one-parent household, and 1 if two-parent
household. We include three measures of household labour market standing by using question
on self-defined current economic status. The variables are organized as binary variables for
no-earner, where none in the household defined themselves as part-time or full-time
employed. One-earner, one of the household members defined themselves as part-time or
full-time employed. Multiple earners, two or more of the household members defined
themselves as part-time or full-time employed. Number of children are included as a
continuous variable measuring number of children below 18 years or below 24 when
economically inactive living in the household.

Macro-level variables
The main source for the macro-level variables is the Social Assistance and Minimum Income
Protection Interim Data-Set (SaMip). SaMip is an on-going research project at the Swedish
Institute for Social Research (Stockholm University). The aim of the data set is to improve the
possibilities to conduct large-scale institutionally informed comparative and longitudinal
analyses of social policy in general and of social assistance and related policy programs of
last-resort in particular. The SaMip contains detailed information regarding the level of
means-tested benefits in 34 industrialized welfare democracies year-by-year for the
period1990-2009. For further information and documentation, see Nelson (2007).

We include five measures of work and family context. Public childcare measures the
enrolment of children below three years of age in public care. Social assistance for lone
parent with two children is the average monthly amounts, excluding housing costs, special
needs benefits and occasional payments. Social assistance for two-parent with two children is
the average monthly amounts, excluding housing costs, special needs benefits and occasional
payments. Unemployment transfers is the unemployment benefits as percentage of GDP.
While it would be optimal to use the unemployment replacement as percentage of own
income, good measures are not available, thus we include Total unemployment rate as well to
get a better or cleaner measure of generosity of unemployment.

Last, we include three measures of structural context. The level of GDP measured in
purchasing power parity EUR. Employment rate female is the percentage of female labour

                                                                                             9
force per working aged citizen. Single household is the percentage of single household in the
country of residence.

Descriptive statistics
Figure 1 indicates that one-parents have significantly higher poverty risk than two-parents, in
all countries. To get a better insight to this problem, table 1 shows descriptive statistics for
some important factors for one-parent and two-parent households. At the individual-level, we
consider several independent variables that are likely to affect poverty. In general, one-parent
households tend to have fewer children compared to two-parent households. Regarding
educational attainment, there are great differences between one-parent and two-parent
households. In all countries, head of one-parent households are more likely to have lower
education compared to head of two-parent household. We found the same figures for higher
education; heads of two-parent households are more likely to have higher education compared
to heads of one-parent households.

                                           (Table 1)

In general, employment is important to mitigate poverty. In all countries, one-parents are
more likely to be non-earner households; nevertheless, cross-country differences are ranging
from less than 2.6 per cent in Denmark to 14.4 per cent in the Netherlands. Conversely, non-
earner households are very rare among two-parent households, and cross-country differences
are smaller than among one-parent households. It is, however, interesting to notice the cross-
country differences in the gap in non-earner status between one-parent and two-parent
households. Surprisingly, the gap between one-parent and two-parent households in non-
employment is nearly 10 percentage points in Sweden, and only 2 percentage points in
Denmark.

To summarize, descriptive statistics confirm that one-parent households are a vulnerable
group. They are more likely to live in poverty, are lower educated, and more likely to be non-
earner households, compared to two-parent households. It seems that one-parenthood entails
difficulties to combine the dual role of being a parent and a worker, leading to less disposable
income compared to two-parent households. In the case of non-employment or low-income,
households have to rely on unemployment or social assistance benefit. Nevertheless, recent
research indicates that those benefits are often inadequate to keep out of poverty (Van
Mechelen et al. 2010; Nelson 2008). The structure and generosity of work-family policies

                                                                                             10
might play an important role explaining cross-country differences in poverty among
households with dependent children.

Table 2 shows descriptive statistics for macro-level variables used in multilevel analyses.
Public childcare varies cross-country, from 2.2 per cent of children below three years in
public childcare in Czech Republic, to 65.7 per cent in Denmark. Family transfers vary from
0.8 per cent of GDP in Poland, to 4.2 per cent of GDP in Denmark. Social assistance
payments (in purchase power parities, PPP) differentiate between one-parent with one child
and two-parent with two children. The minimum income benefit for households with children
varies a lot cross-country. Italy is the only country without any social assistance. The
unemployment benefit, in per cent of GDP, varies from 0.4 per cent in Poland and Romania,
to 3.8 per cent in Belgium.

                                          (Table 2)

To measure the structure of the labour market, we use the total and female employment rate,
and the unemployment rate. The total employment rate varies from 60 per cent in Malta to 80
per cent in Sweden. We find the same pattern for the female employment rate; highest in
Sweden (77 per cent), and lowest in Malta (42 per cent). The unemployment rate is
remarkable high in Spain (20 per cent), and vice versa, very low in Norway (3.5 per cent).

Table 2 indicates differences in generosity of work-family benefits and in labour market
structure. Nevertheless, we ask whether work-family policies can explain differences between
countries in the prevalence of poverty among households. To examine whether the association
between a higher poverty risks at the micro level is spurious or not, we use multivariate
analysis to be able to control for competing explanations.

A multilevel approach
In analyses where the dependent variable is a binary indicator, as here where the dependent
variable is poor or not poor, logistic regression models can be used in order to examine binary
dependent variables. Nevertheless, due to the clustering of households within countries and
the inclusion of country-level variables, standard logistic regression model violates the
assumption of the independence of errors (Rabe-Heseth and Skrondal 2012). A multilevel
modelling technique takes the hierarchical structure of the data explicitly into account and
yields less biased standard errors than a regular logistic regression model (Rabe-Heseth and

                                                                                            11
Skrondal 2012). Consequently, we use multilevel mixed-effects logistic regression model
(estimated in STATA). Mixed logit models predict whether a household is poor based on a set
of individual- and country-level variables. We use the Maximum Likelihood procedure as our
estimation method. When analyzing data with a logistic regression, an equivalent statistic to
R-squared does not exist. The model estimates from a logistic regression are maximum
likelihood estimates arrived at through an iterative process. They are not calculated to
minimize variance, so the OLS approach to goodness-of-fit does not apply. However, to
evaluate the goodness-of-fit of our model we use LogLikelihood and McFadden's pseudo R-
squared. If the predictors in the model are effective, then the penalty will be small relative to
the added information of the predictors. If a model contains predictors that do not add
sufficiently to the model, then the penalty becomes noticeable and McFadden's pseudo R-
squared can decrease with the addition of a predictor. If comparing two models on the same
data, McFadden's would be higher for the model with the greater likelihood. We estimate
several models with stepwise inclusion of variables, first controlling for household
characteristics thus including macro level variables. This would give an insight on how or if
macro level variables mitigate differences between one-parent and two-parent households,
and single-mother and single-father households.

Multilevel results

The partner effect on poverty
Table 3 displays the odds ratios for the first set of models. The baseline model, an empty
model (not shown), analyses the between-country variance without considering any control or
explanatory variables on the individual level. In the first analyses, we calculate the intra class
correlation coefficient (ICC) as 0.09, indicating that country-level differences explain 9 per
cent of the residual variation in poverty risk among families with children.

The first model compares one-parent and two-parent households, with no other control
variables included, and analyses the poverty gap among one-parent and two-parent
households. Model 2 includes only the micro-level variables. We have constructed a reference
group that are comparable with one-parent household. The reference group consists of one-
earner partnered households where head of household has medium education; else, they are
equal on other variables as age, number of children and immigrant status. The reference group
chosen will thus reflect most of the one-parent households and would be comparable with
two-parent households, except that they differ in having a partner.

                                            (Table 3)

                                                                                               12
The estimate from Model 1 shows that one-parent households have much larger odds of being
poor; the factor is 3.1 relative to two-parent households. The poverty gap between one-parent
and two-parent households decreases when controlling for other individual characteristics
(Model 2). The estimates indicate that living in a household with the following characteristics:
one-parent household, no one employed, the head has lower education, a great number of
children and the head is an immigrant, increases one’s odds of being poor. More specifically,
living in one-parent household increases the odds of being poor by a factor of 1.9, relative to
two-parent households when controlling for all the other individual characteristics. According
to the odds ratios, the greatest effects are living in households headed by someone with lower
education where the odds of being poor increases with a factor of 2.8 relative to households
headed by someone with medium education. In contrast, living in households headed by
someone with higher education declines the odds of being poor with a factor of 0.3 relative to
households headed by someone with medium education. Living in no-earner households has a
strong effect of being poor as well, the odds of being poor increase with a factor of 2.5
relative to one-earner households. Alternatively, living in household with multiple earners
reduces the odds of being poor by 0.3 relative to one-earner households. Living in households
where the head is an immigrant increases the odds of being poor with a factor on 2.1 relative
to the reference household. Households with many children have a higher risk of being poor.
This is a continuing variable, which indicates that the effect shown in the table comes from
having one additional child. For each child in the household, the odds of poverty increase by a
factor of 1.8. Age has reducing effects that are more moderate. Living in a household where
head is one year older decline the odds of being poor by a factor of 0.9 relative to household
headed by someone who are one year younger. Age square is positive which indicates that the
reducing effect of age on the poverty risk is decreasing.

In Model 3 we include work and family measures. This model shows that the poverty gap
between one-parent and two-parent households, and the other individual-level variables, are
stable and robust when controlling for work-family measures. Enrolment in public childcare
among children under the age of three is the only work-family variable that is significant at a
5 per cent level. The estimate indicates that a one-unit increase decreases the poverty risk by a
factor of 0.98. The other insignificant results points in the expected direction, except social
assistance among one-parent households. Nevertheless, other variables may confound some of
our work-family variables; Model 4 allows a further inspection.

In Model 4 we include labour market- and country structure. Still, the poverty gap between
one-parent and two-parent households, and the other individual-level variables, is stable and
robust when including country-level variables.

                                                                                              13
Our assumption that labour market- or country structure confounds some work and family
measures variables is supported in Model 4. It is interesting to note that when including
labour market- and country structure variables, some variables turns out to be significant and
vice versa. An increase in the minimum income scheme for two-parent household indicates a
reduction in the odds of being poor by a factor of 0.99 for a one-unit increase in the minimum
income scheme for two-parent household. This seems like a trivial effect; however, it is worth
noticing that the effect shows one unit change. On the other hand, the minimum income
scheme for one-parent households turns out to be significant in Model 4 as well, and one-unit
increase slightly increases the odds of being poor. The direction is unexpected but might be a
results of the disproportion between two-parent households and one-parent households, and
can be understood that the minimum income scheme for one-parent households have no effect
on two-parent households, which are the reference group. To test this assumption, we add an
interaction term between one-parent household and the minimum income scheme for one-
parent households. Because in logistic regression the interpretation of interaction effects is
complicated, we only report the coefficients (not odds ratios) and concentrate on significance
and direction. The interaction term is significant and negative, which indicates that one-unit
increase in the minimum income scheme for one-parent households declines the risk of being
poor among one-parent household (results in Appendix 2). Further, a one-unit increase in the
unemployment benefit declines the odds of being poor by a factor of 0.84 (borderline
significant p
reference household. Two-parent and one-parent households with more than one employed
are statistical different from each other. Living in a one-parent and no-earner household have
the highest odds of being poor, 2.83, relative to two-parents and one-earner households.
Controlling for work-family variables (Model 2 in Table 4) or labour market- and country
structure (Model 3 in Table 4) does not alter any of the odds ratios of the individual variables.
McFadden's pseudo R-squared is decreasing when including work and family measures,
which indicate that the new information do not add sufficiently to Model 3.

                                           (Table 4)

The gender gap in poverty risk among one-parent households
Turning to the gender gap in poverty risk among one-parent households comparing single-
mother and single-father households, we use the same empirical strategy as the analysis in
table 3. The baseline model (not shown), analyses the between-country variance without
considering any control or explanatory variables. The intra class correlation coefficient (ICC)
is calculated as 0.05, indicating that country-level differences explain 5 per cent of the
residual variation in poverty risk among families with children. Model 1 compares single-
mother and single-father households and does not include any control variables, thus
reflecting the gender gap in poverty risk among one-parent households. The estimate from
Model 1 shows that single-father households have significant lower odds of being poor; the
factor is 0.63 relative to single-mother households. In Model 2 (Table 5), individual
characteristics are included. The reference group consists of single-mother, one-earner
households where the head has medium education; else, they are equal on other variables as
age, number of children and immigrant status. The gender gap in one-parent households’
poverty risk is reduced when controlling for other individual characteristics (Model 2). The
odds of being poor are 0.83 if living in single-father households relative to single-mother
households, when controlling for all the other individual characteristics. Other individual
characteristics of interest are that one-parent household where head has higher education has
much lower odds of being poor relative to one-parent household with medium education. On
contrary, the odds of being poor are 2.32 in one-parent household where head has lower
education. Number of children has a very strong effect of the poverty risk, the odds of being
poor increase with a factor of 2.24 for each additional child. Living in no-earner households
has a strong effect of being poor as well, the odds of being poor increase with a factor of 2.54
relative to one-earner households. Alternatively, living in household with multiple earners

                                                                                              15
reduces the odds of being poor by 0.38 relative to one-earner households. Age has moderate
reducing effects. Living in a household where the head is one year older declines the odds of
being poor by a factor of 0.86 relative to household headed by someone who are one year
younger. Age square is positive which indicate that the reducing effect of age on the poverty
risk is decreasing.

                                            (Table 5)

In Model 3, work and family variables are included. The results indicate that the gender gap
among one-parent households does not alter when controlling for work and family measures.
The same pattern is found for the other individual-level variables. Regarding work-family
variables, enrolment in public childcare among children under the age of three is the only
variable that is significant at a 5 per cent level. The estimate indicates that a one-unit increase
decreases the poverty risk by a factor of 0.99 (borderline significant p
whether the association between high poverty risk and parenthood at micro-level is not
spurious, we rely on multivariate analysis to be able to control for competing explanations.

The general picture from the main analysis is that households with multiple earners, head
highly educated, older heads and few children was less likely to be poor, while one-parent
households, non-earners, younger heads, and households with multiple children was more
likely to be poor. Further inspections reveal that individual characteristics shape poverty.
Work-family policies had only limited reducing effects on the poverty gap between one-
parent and two-parent households. Family transfers, social assistance and unemployment
transfers do reduce households’ poverty risk, but individual-level characteristics are
particularly robust and do not alter when controlling for macro-level variables. Moreover, we
carried out a further inspection on how the poverty risks varying after combinations of type of
households and labour market status. One-parent and two-parent in non-earner households are
worse off compared to one-earner or multiple-earner households, this result were expected.
However, family structure does not matter between non-earner households; one-parent and
two-parent households where not statistical different from each other. On the other hand,
family structure does matter in one-earners and multiple-earner households. Among one-
earner households, differences in the poverty gap might be results of differences in one-parent
households’ individual factors that are not encompassed by the models, and that work-family
policies are not sufficient to tighten the poverty gap. Further, one-parents and two-parents
from multiple-earner households are statistical different from each other. We may explain this
finding by the composition of the employed in the household. In one-parent households,
youngsters are more likely to be an additional earner, which implies that they typically work
part-time or in low-wage jobs. We restricted our sample to households with children, where a
child was defined as a person below 18 years or below 24 when economically inactive living
in the household, however, it is possible that one-parent households have economic active
children in addition to economic inactive children living in the household.

In the last analyses, we examined whether work-family policies reduce the gender gap in
poverty among one-parent households. The results indicate that single-mother households had
a significant higher poverty risk compared to single-father households. Moreover, the same
picture arises, individual characteristics shape the poverty risk among one-parent households,
and including work-family characteristics do not diminish the gender gap. In contrast to the
joint analyses of one-parent and two-parent households, it seems that individual
characteristics count more for one-parent households though none of the labour market- or
country structure variables was significant.

Lastly, did our results give an answer to our research questions? To sum up, our first research
question was; can work-family policies reduce the poverty gap between one-parent and two-

                                                                                            17
parent households? The second was; can work-family policies reduce the gender gap in
poverty among single-mothers and single-father households? Overall, the results indicate that
work-family policies are not sufficient to reduce the poverty gap between one-parent and two-
parent households, or between single-mother and single-father households. Individual
characteristics were the main factor to tighten the poverty gap. Obviously, some work-family
measures do reduce households’ poverty risk although the effect on a reduction of the poverty
gap was very limited. For instance, high enrolment in public childcare indicates a lower
poverty risk, which points in the direction that policy which arranges possibilities for
combining work and family life carry out lover poverty risk among households with children.
In contrast, high female labour force participation is not significant but points in the directions
of higher poverty risk. At first sight this might be regarded as somewhat surprising. A
probable explanation is that high female labour force participation pushes the median income
upwards. Consequently; high female labour force participation gives a higher poverty line.
Vulnerable groups outside the labour market thus would have relatively higher poverty risk
compared to those employed. This is supported by the strong effect of household-level
employment found in our analyses.

It is worth noticing that our definition of poverty is based on the equivalent household income
for one year, and this might underestimate the effect of work-family policies on households’
poverty risks. Measuring poverty only for one year would probably include many households
hit by “random” shocks in their household economy (e.g. after a divorce it takes some time
before a household would receive benefits), thus the benefits are not included in the income
measure used in EU-SILC (especially for those countries using administrative register data).
Partly to address the limitations of this study, we encourage future research to give more
effort on studying the effect of work-family policies on long-term poverty, on individual-level
employment, and the use of public childcare among vulnerable groups. This is essential to
extend the understanding of vulnerable groups’ use of work and family friendly measures,
which in turn might be crucial to understand the effect of work-family policies on the poverty
risk among household with children.

                                                                                                18
References
Bradshaw, J. (2006). Child benefit packages in 15 countries in 2004. In Children, Changing
     Families and Welfare States, edited by J. Lewis. Cheltenham: Edward Elgar.
Campione, W. (2008). Employed women’s well-being: The global and daily impact of work.
    Journal of Family and Economic Issues, 29, 346-361.
Casper, L. M., McLanahan, S. S., and Garfinkel, I. (1994). The Gender-Poverty Gap: What
     We Can Learn from Other Countries. American Sociological Review, 59(4); 594-605.
Christopher, K. 2002. Single Motherhood, Employment, or Social Assistance: Why are U.S.
      Women Poorer than Women in Other Affluent Nations? Journal of Poverty, 6 (2):61 -
      80.
Christopher, K. 2002. Welfare State Regimes and Mothers' Poverty. Social Politics: International
      Studies in Gender State and Society, 9 (1):60-86.
Christopher, K. 2005. A “Pauperization of Motherhood”? Single Motherhood and Women's
      Poverty over Time. Journal of Poverty 9 (3):1-23.
Christopher, K., England, P., Smeeding, T.M. and K. R. Phillips. 2002. The gender gap in poverty
      in modern nations: Single motherhood, the market, and the state. Sociological Perspectives
      45 (3):219-242.
Ciabattari, T. (2007). Single mothers, social capital, and work-family conflict. Journal of
     Family and Economic Issues, 28, 34-60.
Daly, M., and Jane Lewis. 2000. The concept of social care and the analysis of contemporary
      welfare states. British Journal of Sociology 51 (2):281-298.
DeBord, K., Canu, R. F., and Kerpelman, J. (2000). Understanding a work-family fit for
    single parents moving from welfare to work. Social Work, 4, 313-324.
Edin, K., and Kefalas, M. (2005). Promises I can keep: Why poor women put parenthood
      before marriage. Berkeley: University of California Press.
European Commission (2002). Joint Employment Report Luxembourg, Office for Official
     Publication of the European Communities.
Esping-Andersen, G. (1990). The Three Worlds of Welfare Capitalism. Cambridge: Polity Press.
Esping-Andersen, G. (1999). Social Foundations of Postindustrial Economies. Oxford, Oxford
     University Press.
Esping-Andersen, G. (with D. Gallie, A. Hemerijck, and J. Myles) (2002). Why We Need a
     New Welfare State. Oxford: Oxford University Press.
Ferrarini, T. (2006). Families, states and labour markets. Cheltenham: Edward Elgar.
Forma, P. (2009). Work, family and intensions to withdraw from the workplace. International
    Journal of Social Welfare, 18, 183-192.

                                                                                             19
Fraser, N. (1994). After the Family Wage: Gender Equity and the Welfare State. Political
      Theory 22 (4):591-618.
Gemelli, M. (2008). Understanding the complexity of attitudes of low-income mothers toward
    work and family in the age of welfare reform. Gender Issues, 25, 101-113.
Ghysels, J. and Van Lancker, W. (2011). The unequal benefits of activation: an analysis of the
     social distribution of family policy among families with young children. Journal of
     European Social Policy 21(5): 472-485.
Haddock, S. A., Zimmermann, T. S., Ziemba, S. J. and Lyness, K. P. (2006). Practices of dual
     earner couples successfully balancing work and family. Journal of Family and Economic
     Issues, 27, 207-234.
Huang, Y. H., Hammer, L. B., Neal, M. B. and Perrin, N. A. (2004). The relationship between
    work-to-family conflict and family-to work conflict: A longitudinal study. Journal of
    Family and Economic Issues, 25, 79-100.
Howe, B., and Pidwell, R. (2004). Single parents and paid work. Australian Journal of Social
    Issues, 39, 169-181.
Lewis, J. (1992). Gender and the development of welfare regimes. Journal of European
     Social Policy 2 (3):159-173.
Lewis, J., Knijn, K., Martin, C. and Ostner, I. (2008). Patterns of Development in Work/Family
     Reconciliation Policies for Parents in France, Germany, the Netherlands, and the UK in the
     2000s. Social Politics: International Studies in Gender Stat and Society 15 (3):261-286.
Malone, K., Stewart, S. D., Wilson, J., and Korsching, P.F (2010). Perceptions of financial
     well-being among American women in diverse families. Journal of Family and
     Economic Issues, 31, 63-81.
Mannis, V. S. (1999). Single mothers by choice. Family Relations, 48, 121-128.
Misra, J, Budig, M. J. and Moller, S. (2007a). Reconciliation policies and the effects of mother-
     hood on employment, earnings, and poverty. Journal of Comparative Policy Analysis 9
     (2):135-55
Misra, J, Moller, S and Budig, M. J. (2007b). Work Family Policies and Poverty for Partnered
     and Single Women in Europe and North America. Gender and Society, 21; 804-825.
Moen, P., and Roehling, P. (2005). The career mystique. Lanham, MD: Rowman and Littlefield.
Montanari, I. (2000). From family wage to marriage subsidy and child benefits: controversy
    and consensus in the development of family support. Journal of European Social Policy
    10 (4):307-333.
Neblett, N. G. (2007). Patterns of single mothers’ work and welfare use. Journal of Family
     and Economic Issues, 28, 1083-1112.
Nelson, K. (2007). Introducing SaMip: The Social Assistance and Minimum Income Protection
     Interim Dataset. S-WoPEc No. 11/2007. Swedish Institute for Social Research.

                                                                                              20
Nelson, Kenneth. 2008. Minimum Income Protection and European Integration: Trends and
     Levels of Minimum Benefits in Comparative Perspective 1990-2005. International
     Journal of Health Services 38 (1):103-124.
O'Connor, J, Orloff, A. S, and Shaver. S. (1999). States, Markets, Families. Cambridge:
     Cambridge University Press.
O’Connor, J. (2005). Employment-anchored social policy, gender equality and the open
    method of policy coordination in the European union. European Societies 7 (1):27-52.
Orloff, A. S. (1993). Gender and the Social Rights of Citizenship: The Comparative Analysis
      of Gender Relations and Welfare States. American Sociological Review 58 (3):303-328.
Rabe-Hesketh, S. and Skrondal, A.(2008). Multilevel and Longitudinal Modeling Using Stata.
     Stata Press, Second Edition.
Schieman, S., and Young, M. (2011). Economic hardship and family-to-work conflict: The
     importance of gender and work conditions. Journal of Family and Economic Issues, 32, 46-
     61.
Son, S., and Bauer, J. W. (2010). Employed rural, low-income, single mothers’ family and work
      over time. Journal of Family and Economic Issues, 31, 107-120.
Tuttle, R., and Garr, M. (2009). Self-employment, work-family fit, and mental health among
      female workers. Journal of Family and Economic Issues, 30, 282-292.
Van Lancker, W. (2012). "The European World of Temporary Employment: Gendered and
     Poor?" European Societies 14(1): 83-111.
Van Lancker, W., Ghysels, J., and Cantillon, B. (2012). An international comparison of the
     impact of child benefits on poverty outcomes for single mothers. CSB Working Paper
     No. 12/03. Antwerp: Herman Deleeck Centre for Social Policy
Van Mechelen, N., Marx, I., Marchal, S., Goedemé, T., and Cantillon, B. (2010). The Evolution
     of Social Assistance and Minimum Wages in 25 European Countries, 2001 2009. Antwerp:
     Herman Deleeck Centre for Social Policy.
Winslow, S. (2005). Work-family conflict, gender, and parenthood, 1977-1997. Journal of
     Family and Economic Issues, 26, 727-755.

                                                                                          21
Tables
Table 1. Descriptive statistic of the analytical sample, individual level

                                                Number of                         Head low        Head meadium
                       N            Age          children        Non-earners      education         education     Head higher education

                                  one-   two-   one-     two-     one-   two-     one-     two-     one-     two-
                                parent parent parent   parent   parent parent   parent   parent   parent   parent one-parent two-parent
                          All      HH     HH     HH       HH       HH     HH       HH       HH       HH       HH         HH         HH
AT   Austria            1781     40,17 41,98    1,48      1,8      7,9      1     14,6      3,9     67,7     59,6       17,7       36,4
BE   Belgium            2023     41,12 41,48    1,66     1,91      9,9    2,3     30,5      8,9     37,6     31,9       31,9       59,1
BG   Bulgaria           1463     42,77 42.68    1,41     1,59      3,8    1,1     17,3     15,7       50     49,8       32,7       34,5
CZ   Czech republic     2637     39,81 41,06    1,59     1,69      5,9    0,4      5,9      1,6     78,2     70,7       15,8       27,8
DE   Deutschland        3175     41,95 42,53    1,39     1,72     11,1      1     12,2      2,4     55,6     39,4       32,3       58,2
DK   Denmark            2154     43,25 42,72    1,57     1,89      2,6    0,4     16,2      3,7     47,8     39,1       35,9       57,1
EE   Estonia            1804     42,51 42,08    1,71     1,85      8,5    1,4        8      3,1       54     48,9       37,9       47,9
ES   Espania            4659     42,26 43,46    1,53     1,67      6,7    2,3     34,2     23,5     28,6     26,7       37,2       49,8
FI   Finland            3600     43,68 42,83    1,65     1,98      6,5      1     12,8      2,4     42,4     35,3       44,8       62,3
FR   France             3652     42,31 41,88     1,7     1,93      7,4    1,6     17,1      5,3     50,5     43,9       32,5       50,8
HU   Hungary            3266     42,18 42,49    1,58     1,79        9    2,4     12,1        8     58,4       59       29,6       32,9
IT   Italia             5755     42,36 43,97     1,5     1,62      5,2    1,9     24,5     17,2     58,2     57,5       17,3       25,3
LT   Lithuania          1538     42,66 43,49    1,45     1,61     10,6    2,4      3,7      2,5     55,6     50,9       40,7       46,6
LU   Luxenburg          2094     42,55 42,38    1,71     1,92      9,8    1,7     36,3     23,9     37,2     39,7       26,5       36,3
LV   Latvia             1805     41,24 41,85    1,49     1,66      5,9    1,8     10,1      4,9     58,7     53,1       30,5         42
MT   Malta              1272     41,43 43,86    1,72     1,73       17    2,1       71     38,9       21     34,9          8       26,1
NL   Netherland         3554     44,42 43,53    1,66     1,95     14,4    1,8     14,9      4,5     45,9     40,3       39,1       55,2
NO   Norway             1996     42,19 42,75     1,6     1,97      5,5      1     16,2      4,9     40,8     35,8         43       59,3
PL   Poland             4659     42,68 42,61    1,65     1,81       12    1,8     11,2      2,2     63,9     67,4       24,9       30,4
PT   Portugal           1532     42,24 43,69    1,65      1,6      4,1    1,4       52     46,2     29,7     31,6       18,2       22,2
RO   Romania            1812     43,54   43,6   1,34     1,55      6,9    2,1     15,5     10,8     62,9     70,8       31,6       18,4
SE   Sweden             2462      43,6 41,31    1,59     1,87     10,6      1     12,1      1,4     54,9     44,4       32,9       54,1
SI   Slovenia           3902     42,03   43,9   1,58     1,76      6,5    1,2      9,5      2,4     59,7     57,3       30,7       40,3
SK   Slovak republic    2131     43,24 43,84    1,54     1,85      8,6      1      2,6        1     64,9     58,1       32,5       40,9
UK   United Kingdom     2228      39,3 42,97    1,81     1,87      5,6      1     17,2      5,4     61,9     42,1       20,9       52,5

                                                                                                                                     22
Table 2. Descriptive statistic of the analytical sample, country level
                                    Public         Family      Social        Social    Unemploy-       Total                             Single
                                child-care      transfers  assistance    assistance ment benefit employment Employment      Unemploy- household
                         N
Table 3.Multilevel results, all. OR

                                       Model 1   Model 2     Model 3   Model 4

Household characteristics
One-parent                             3,10***     1,93***   1,93***   1,93***
Age                                                0,89***   0,89***   0,88***
Age squared                                        1,12***   1,12***   1,12***
Number of children                                 1,82***   1,82***   1,81***
Low education                                      2,78***   2,79***   2,78***
Medium education (ref)
High education                                     0,31***   0,31***   0,32***
Non-earners                                        2,31***   2,31***   2,31***
One-earner (ref)
Multiple-earners                                   0,28***   0,28***   0,28***
Immigrants                                         2,10**    2,10***   2,11***
Work and family
Public childcare                                             0,98***   0,99
Family transfers                                             0,930     0,97
Social assistance (single hh)                                1,000     1,01**
Social assistance (partnered hh)                             0,990     0,99**
Unemployment transfers                                       0,820     0,84
Unemployment rate (total)                                    1,07***   1,04
Labour market structure
Female labour force participation                                      0,99
Country structure
GDP 2010                                                               0,99
Single household                                                       0,98
Constant                               0,15***               3,46***   3,54***

ICC                                     0,091     0,085      0,031      0,027
ll(model)                              -27734    -23680      -23666    -23665
McFadden pseudo-R2                      0,032     0,035      0,012      0,012
AIC                                     55474     47382       47367     47368
BIC                                     55502     47482       47521     47560

Note: p
Table 4. Multilevel results, combination of type of household and employment status.
OR.

                                              Model 1       Model 2       Model 3

Household characteristics
One-parent household and:
Non-earners                                    2,83***       2,82***        2,83***
One-earners                                    1,26***       1,26***        1,27***
Multiple-earners                               0,62***       0,62***        0,62***
Two-parent household and:
Non-earners                                    2,47***       2,47***        2,47***
One-earners (ref)
Multiple-earners                               0,23***       0,23***        0,24***
Head of household
Age                                            0,90***       0,89***        0,89***
Age squared                                    1,11***       1,11***        1,11***
Number of children                             1,82***       1,81***        1,82***
Low education                                  2,78***       2,79***        2,79***
Medium education (ref)
High education                                 0,32***       0,31***        0,32***
Immigrants                                     2,11***       2,11***        2,12***
Work and family
Public childcare                                             0,98**         0,99
Social assistance (single hh)                                1,010          1,01*
Social assistance (partnered hh)                             0,990          0,99*
Unemployment transfers                                       0,960          0,81*
Labour market structure
Female labour force participation                                           0,99
Country structure
GDP 2010                                                                    0,99
Single household rate                                                       0,98
Constant                                       2,71***       3,52***        3,56***

ICC                                                 0,086         0,031         0,011
ll(model)                                         -23609        -23601        -23594
McFadden pseudo-R2                                  0,035         0,011         0,010
AIC                                                47244         47239         47232
BIC                                                47362         47402         47432

Note: p
You can also read